AI Agent on SAP BTP, the chatbot for smarter business processes

The integration of Artificial Intelligence into business processes is accelerating significantly, but the real leap forward happens when AI is designed as an agent—capable not only of understanding but also of acting within systems. In this context, Mashfrog has developed an intelligent chatbot based on SAP AI Core, connecting generative models, OData services, and document repositories such as SharePoint to provide users with accurate, actionable responses to their requests.

A concrete application of the Agentic AI approach and of the Group’s technological expertise, particularly within the SAP ecosystem.

Simplifying access to SAP processes

In the day-to-day management of business processes, users often need to retrieve information scattered across multiple systems: documents, application services, SAP modules. Repetitive operations—such as checking the status of a purchase request or reviewing maintenance data—require time and specific expertise.

Organizations are seeking intuitive solutions that allow anyone—even without technical knowledge—to interact with digital processes through natural language. The goal: increasing efficiency, reducing errors, and freeing up time for higher-value activities.

AI Agent chatbot on SAP AI Core

Mashfrog’s solution is built on SAP AI Core, the SAP Business Technology Platform service designed to manage AI workloads in a scalable and governed way. The solution integrates three key pillars:

1. Connection to advanced AI models
   SAP AI Core makes it possible to orchestrate multiple Large Language Models (OpenAI, Gemini, Mistral, and others), selecting the one best suited for the task at hand.
    
2. Direct access to corporate content
   Thanks to native integration with SharePoint, the chatbot can retrieve information from stored documents, expanding its “knowledge” beyond what it was originally trained on.
    
3. Integration with SAP OData services
   The true agentic value lies in the chatbot’s ability to:

• understand the user’s request;
• determine whether an OData service exists to fulfill it;
• autonomously generate an OData query;
• send it to the SAP system;
• receive the JSON response and translate it into natural language.

In this way, the AI doesn’t just provide textual information — it truly interacts with corporate systems, performing actions on demand.

The AI chatbot has already been successfully and fully implemented for purchase requests (PRs): users can request information about their assigned PRs, check their status, release them, or retrieve updated data — all without accessing SAP menus.

More efficiency, less complexity, more value from data

Adopting an agentic chatbot brings concrete, measurable benefits:

Immediate access to SAP processes
Users interact in natural language and receive accurate responses without navigating transactions or applications.

Reduction of operational time
Tasks such as consulting, checking, or releasing data become instantaneous.

Democratization of information
Even non-technical users can access complex SAP content.

Quality and consistency of responses
Thanks to the combined use of SharePoint documentation and reliable language models.

Security and governance
SAP AI Core ensures a controlled environment fully aligned with corporate policies.

Applicability and replicability within the corporate workflow

One of the key strengths of the solution is its scalability: once the agentic architecture is defined, the chatbot can be connected to any available OData service.

Examples of possible extensions include:

• Plant Maintenance: querying maintenance notifications, requesting the status of technical orders, retrieving information about a technical location.
• Sales & Distribution: retrieving customer order status, item availability, and information on deliveries and shipments.
• Human Resources: viewing personal data, vacation balances, certifications, and HR documents.
• Finance & Controlling: accessing reports, monitoring costs, and checking accounting entries.
• Supply Chain Management: immediate visibility into stock levels, requirements, and material movements.

Thanks to the modular nature of the architecture and the combined power of agentic AI and SAP, the solution can rapidly evolve according to the company’s needs, progressively expanding its functional scope.

Replicability is maximized: simply expose new OData services or integrate new document sources, and the chatbot can immediately extend its operational capabilities.

The developed solution demonstrates how AI designed with an agentic approach becomes a true operational assistant capable of interacting with SAP processes.

Mashfrog enhances this vision through its solid expertise in SAP BTP, cloud architectures, OData integration, and the orchestration of advanced AI models. The result is a versatile platform capable of evolving and adapting to business needs, accelerating digital transformation in a concrete and measurable way.